3DMark05 WR -> 34634

This was achieved during last weekends Assembly... cool, literally:

-Intel Core 2, 2,93 GHz -> 5,341 GHz
-ATI Radeon HD 2900 XT
-Asus P5W64 WS Pro
-Microsoft Windows XP

...And lot's of liquid nitrogen, of course :)

Compare URL:

http://service.futuremark.com/compare?3dm05=3316824

Imagine what it coulda done with a real video card like an 8800 Ultra. :)

Fanboism, how cute!

heh... I was just razzing a bit... BUT right now the 8800 ultra is the one to beat. I've owned ati and I've owned nvidia. Whichever is the best at the time I'm goin to buy, is what I buy. Same with processors. I'm probably going back to intel my next upgrade because they win... hands down right now as a matter of fact. I've been AMD-happy for the last 4-5 years, but they aren't pulling their weight right now.

No fanboism at all... Just some good natured ribbing of our local mustang pilot. :)

SPEAKING OF... Yo Mustang... You gonna be at Rickenbacker for the Gathering of Mustangs 2007 this year? (Sep 27-30 I think).

If so I'm SERIOUSLY thinking of heading out that direction come fall.

But, Kanati, there are things to consider about the ATI card in question. With the right PSU, and air cooling (yes...), the GPU can be overclocked to 1ghz. With water cooling, a little more. Also the card can be in SLI (crossfire), and the performance of an ATI in crossifre usually is very good, compared to the Nvidias in SLIe.

So, overclocked, it is still a very close call as to wether it's 8800 GTX / Ultra or 2900XT in SLI / crossfire. Many serious system builders are considering the 2900XT in crossfire and overclock that with a 1000+ watts PSU.

As for the CPU, of course Core 2 is beating AMD hands down. So, we have to see if the so called Phenom is any good. The only good the AMD is, IMHO is price. They have the low cost market share since last winter with their massive price cuts. And as usual, Intel keeps their dominating price way above. It's like we're back in the days where AMD didn't have a competing product in the high end.

I will have to take a look at 2900XT in crossfire, but I have generally decided across the board not to OC anything I own. Stability has been enough of an issue that I just don't like fighting with it.

And it is indeed a shame about the AMDs. I gave some thought to going Dual FX74 for a bit but the benchmarks on it weren't that great. And there's only one board that allows dual fx as well.

Assuming I don't see something stupendous with the 2900 x2, I'll probably go 8800ultra with a 6850 core2duo. Two cores are good enough, and obviously from mustang's score there, it's seriously overclockable if I can get over my trepidation of doing so.

In all honesty, the big advantage Nvidia has with the 8000 series is the fact they can clock the shaders differently than the whole GPU. So in fact you can have the GPU at 600 mhz, but the shaders at 1.5 ghz. That is where the 8800 series for example win a lot. And that is why this margin shrinks when you overclock the HD2900 XT of the ATI, since when you overclock the ATI, you also overclock the 320 shader cores, so in essence, when you overclock the ATI to 1 Ghz, you also have all the shader cores running at 1 ghz, and that's a gigantic leap from the base 650 / 675 mhz. Also, the memory interface of the R600 is 512 bit, as opposed to 384 bit for the 8000GTX / Ultra, which enables up to 150 gigs/sec.

Finally, the board of the HD2900XT has 7 digital (PWM), which enables it to totally control the power usage from software. It's the very first graphics card to include that, and used to be the thing that kept all graphics cards from perfoming some serious overclocks.

http://www.xtremesystems.org/forums/showthread.php?t=144198

Nope, sadly... that would rock!

Sorry, i see less and less reason for an app like 3dmark. It was fun to watch it in the beginning and do some curious tests beside of seeing it like a demo but as the gap is getting wider and wider and they don't want to offer a benchmark which is useable on old slow cards as well as on new high end cards it just doesn't make sense anymore to me. Pyramid application...

Every new 3DMark is meant for benchmarking the latest (at the time of releasing) and coming cards - NOT the old ones. Older 3DMarks exist for that purpose.

It's not one product for whole range of VooDoos to GF10990 at all, and never has been. You should always use right 3DMark for benchmarking the right gen cards (SM/DX).

And we (futuremark) still see a reason to make 3DMarks, as do all major 3D card makers and press... and few million users that download and buy the product(s) every year.

We had this discussion before and i stand with the opinion that 3dmark should consist of two attractive parts: a) using new hardware stuff (although also keeping here in mind that this runs not only on the latest highend cards in resonable framerates) and b) doing tests with which you can compare older and newer cards.

This isn't possible with the current concept and as you once said the numbers from older and newer tests aren't compareable. This makes it simply unattractive to me as i primary need to know how fast certain things run on different hardware.

I see a resason for the benchmark for the fanboys who like to state with their hardware or the sponsored ones who get hardware from different vendors. If the mags would stop using the tool, the interest would decrease very soon too.

I don't know if there is interest in 3dmark from the vendors beside of having to follow this vicious circle and how much they benefit from the tool in terms of comparing their latest technologie with the competition or squeezed out bugs this way.

I would prefer if they would enhance their drivers overall much better. I would welcome a year (or whatever it takes) of no new hardware features in cards beside of bugfixing, improving already built in concepts, less heat due to die shrinks or improved layouts, ... and providing more mature drivers instead.

What's the framerate difference you get in 3dmark for the latest highend and lowend solution? I bet 3dmark isn't enjoyable at all when you're looking at the latest lowend.

That kind of comparison isn't possible, I think... 3DMark requires certain tech level (SMs) to even start and low end cards don't have them. Also lowend cards usually have less memory (because it costs) than required and swapping would make results flawed.

And of course it isn't enjoyable for ANYONE [when the new 3DMark is released] because it has be heavy enough to stay usable tool for the next ~2 years it usually takes between DXses.

But how reasonable is it providing a benchmark which isn't able to provide compareable and enjoyable results even with the latest generation of cards which are featurewise the same (leaving the video functionality out)? If were taking nVIDIA ranging from a 8400GS to a 8800Ultra for instance?

Don't missunderstand me, my primary goal isn't to bash your product but i just came the way along from "wow this looks cool" over "well i need a better card" to "it's simply useless to me". I also miss different aspects like how much stressing a certain GPU increases the heat of it. I just don't get information out of it which is important to me nowadays.

It's not a screen saver. It's not supposed to be enjoyable. The fact that it CAN be enjoyable is a bonus. It's supposed to give comparable results of current and future generation equipment. Numbers. Eye candy is tertiary at best.

Oh really, is this the case?

I wonder why they don't just throw some randomly generated meshes and textures at the user then and use art and eye candy instead...hmm maybe because it sells better this way?!

Future generation != future highend cards only, the benchmark ignores this fact.

By the way how about Cinebench? http://www.maxon.net/pages/download/cinebench_e.html

http://www.xtremesystems.org/forums/showthread.php?t=156596

Kinc and Shamino breaks 27,000 in 3DMark 06 with Mushkin and ASUS HD 2900 XT CrossFire!

Cool...